Discussion Intel current and future Lakes & Rapids thread

Page 88 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
You're so funny.
AMD had the tools to get ahead. They had an EDA lead as shown by Excavator. It wasn't till Cannonlake/Sunnycove that Intel did the same. So, they had bad foresight and squandered their advantage. It isn't funny business. It is all hindsight of being an individual in 2019, knowing the performance of processes in 2013(AMD's 22nm node) and 2015(AMD's 14nm node).

Skylake-AVX512 -> Sunnycove 14nm is a track height reduction. Much like Steamroller 13T to Excavator 9T.
Which would be 18 mm squared(Skylake AVX512) to something above 12 mm squared(Sunnycove 14nm). *shrug*
Someone post Sunnycove microarchitecture/Whitley platform:{Cooperlake - Icelake}
 
Last edited:

Thunder 57

Platinum Member
Aug 19, 2007
2,675
3,801
136
AMD had the tools to get ahead. They had an EDA lead as shown by Excavator. It wasn't till Cannonlake/Sunnycove that Intel did the same. So, they had bad foresight and squandered their advantage. It isn't funny business. It is all hindsight of being an individual in 2019, knowing the performance of processes in 2013(AMD's 22nm node) and 2015(AMD's 14nm node).

Skylake-AVX512 -> Sunnycove 14nm is a track height reduction. Much like Steamroller 13T to Excavator 9T.
Which would be 18 mm squared(Skylake AVX512) to something above 12 mm squared(Sunnycove 14nm). *shrug*
Someone post Sunnycove microarchitecture/Whitley platform:{Cooperlake - Icelake}

You have no idea what you are talking about, and you claim there will be AMD products on FDX nodes that will never happen. You seem to have an affection for the Con cores, which is fine. There is a reason AMD did a total new design though, the Con cores were never going to deliver. I have no idea why you freely post blatant nonsense. AMD did a great job with Steamroller and Excavator, getting power down to such levels on a power hungry architecture. But it was clear that a new design was needed, and AMD has done rather well with Zen.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
AMD has done rather well with Zen.
AMD did great with Excavator, but they did poorly with Zen.
AMD did a great job with Steamroller and Excavator, getting power down to such levels on a power hungry architecture.
All Bulldozer cores released are derived from the 5 GHz Opteron. It is a big core, with a huge TDP budget. Intel is having the same issues with Skylake now.
There is a reason AMD did a total new design though, the Con cores were never going to deliver.
All designs after 32nm/28nm are total, complete, full new designs with no IP re-usage. 22nm Steamroller != 28nm Steamroller, they are two different architectures with the same name. Much like how Intel calls the Cannonlake core, Palm Cove to give it a semblance to the new core. Cannonlake is not a Cove core, it is a Lake core.
You have no idea what you are talking about, and you claim there will be AMD products on FDX nodes that will never happen.
Early GlobalFoundries does things for one reason and that is satisfying AMD. GlobalFoundries licensing STMicroelectronics' FDSOI nodes is the same as AMD developing products for said nodes. Which they have, and which some will be released. AMD and Mubadala Technology share a strict AMD-focused relationship.
 

Thunder 57

Platinum Member
Aug 19, 2007
2,675
3,801
136
AMD did great with Excavator, but they did poorly with Zen.

You are so wrong. AMD has done nothing but gain market share since Zen has launched.


All Bulldozer cores released are derived from the 5 GHz Opteron. It is a big core, with a huge TDP budget. Intel is having the same issues with Skylake now.

Source? We know AMD expected higher clocks but even that was not enough to be competitive.


All designs after 32nm/28nm are total, complete, full new designs with no IP re-usage. 22nm Steamroller != 28nm Steamroller, they are two different architectures with the same name. Much like how Intel calls the Cannonlake core, Palm Cove to give it a semblance to the new core. Cannonlake is not a Cove core, it is a Lake core.

I have no idea what you are talking about here. There was no 22nm Steamroller.

Early GlobalFoundries does things for one reason and that is satisfying AMD. GlobalFoundries licensing STMicroelectronics' FDSOI nodes is the same as AMD developing products for said nodes. Which they have, and which some will be released. AMD and Mubadala Technology share a strict AMD-focused relationship.

Right, have you not heard about Mubadala selling their shares in AMD and GF looking for a buyer?
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
Source? We know AMD expected higher clocks but even that was not enough to be competitive.
It doesn't matter the 5 GHz Opteron was a 4-wide integer CMP core processor on the 65nm PDSOI node. *sarcasm*K8 with its 6-wide integer CMP core processor is better */sarcasm*.
I have no idea what you are talking about here. There was no 22nm Steamroller.
Steamroller within the years 2009-2012 was a 22nm core. While profiles state 22PDSOI, AMD was already sending memos to ATIC/GlobalFoundries to drop 22PDSOI from the roadmap.
Right, have you not heard about Mubadala selling their shares in AMD and GF looking for a buyer?
They are selling ~35 million, to get ~75 million shares. Mubadala is not selling GlobalFoundries. They might sell some Fabs, but Globalfoundries is here to stay.

---
If people don't want me to go on about this stuff. Do not troll me or at least don't reply to me.
 

DrMrLordX

Lifer
Apr 27, 2000
21,633
10,845
136
If people don't want me to go on about this stuff. Do not troll me or at least don't reply to me.

Is that a promise?

Back on topic:

If we're seeing IceLake laptops by June, then should we expect IceLake-S anytime soon? Can't happen fast enough for Intel.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
If we're seeing IceLake laptops by June, then should we expect IceLake-S anytime soon? Can't happen fast enough for Intel.
I think that is after Comet Lake-S. Since, the octo-core CNL-S was a no go.

Z390 would I guess support 8C/16T and 10C/20T.
Then, a new chipset to support Icelake-S and Tigerlake-S. To be competitive, lets assume it will be the Z590.

Isn't there going to be a new platform to support EMIB(+successor) with HBM2e/3. For the new C-suffix designs?
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,633
10,845
136
I think that is after Comet Lake-S.

I'm kinda hoping Intel manages to get their act together, ramp up production of 10nm, and avoid another band-aid release like Comet Lake-S. Releasing that chip after Zen2 hits the streets will be a disaster.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
I'm kinda hoping Intel manages to get their act together, ramp up production of 10nm, and avoid another band-aid release like Comet Lake-S. Releasing that chip after Zen2 hits the streets will be a disaster.
The most aggressive roadmap I can create from internet-derived sources.

Comet Lake-S for 2019 -> "14nm++" P1272.XX -> Skylake / G9.5
Tigerlake-S for 2020 -> "10nm++" / 1274.12 -> Willowcove / G12
Alderlake-S for 2021 -> "7nm" / 1276 -> Goldencove / G13
Meteorlake-S for 2022 -> "5nm" / 1278 -> Oceancove / G14

Phoenix Technologies in BIOS development has CometLake at queue position 8, and TigerLake at queue position 7.
 
Last edited:

jpiniero

Lifer
Oct 1, 2010
14,599
5,218
136
I'm kinda hoping Intel manages to get their act together, ramp up production of 10nm, and avoid another band-aid release like Comet Lake-S. Releasing that chip after Zen2 hits the streets will be a disaster.

As long as Comet Lake is still faster in games, it'll do fine. HEDT is screwed but there's not much they can do about it.
 
  • Like
Reactions: ozzy702

Thunder 57

Platinum Member
Aug 19, 2007
2,675
3,801
136
If people don't want me to go on about this stuff. Do not troll me or at least don't reply to me.

I am not trolling you. If anything, you are the troll. You say AMD did great with Excavator, but poorly with Zen. That is objectively false. In the interest of keeping this on topic, that is the last I will say on the matter.


As long as Comet Lake is still faster in games, it'll do fine. HEDT is screwed but there's not much they can do about it.

Probably. But if AMD can get the frequency boost they need with 7nm, along with the IPC improvements, it may come down to pricing.
 

jpiniero

Lifer
Oct 1, 2010
14,599
5,218
136
By the way I have a feeling that the XPS model mentioned at CES might be the only Dell model offers with Icelake, similar to how Lenovo only have those two models.

Maybe HP then gets the cut down i3/Celeron/Pentium chips for a model or two? I bet there will be disabled GPU chips as well.
 

DrMrLordX

Lifer
Apr 27, 2000
21,633
10,845
136
We finally have a 4.0Ghz Pentium...

Wow no turbo? Interesting. Not much there to see otherwise. How much longer does Intel want to/need to push out 14nm++ Pentiums and such? It is my understanding that they prioritized much of their 14nm++ wafer allocation to larger dice for more-expensive CPUs. It also seems like they've been able to do a 2c/4t chip on 10nm for awhile now, and that they might actually get some decent ones with working iGPUs by this July or so. You would think now would be the time for them to make the transition of their Pentium lineup to 10nm.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
The last 4.0 Pentium was the cancelled P4 580 from 2004.
There was also a Prescott P4 4.0 that never saw the light of day from 2005.
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,633
10,845
136
The last 4.0 Pentium was the cancelled P4 580 from 2004.
There was also a Prescott P4 4.0 that never saw the light of day from 2005.

Huh, thought Cedar Mill got up there, but I see that it never did. The fastest was the 661 @ 3.6 GHz.
 

coercitiv

Diamond Member
Jan 24, 2014
6,203
11,909
136
Wow no turbo? Interesting.
Why have turbo when the chip won't reach power limits anyway? You clock as high as it makes sense and that's it.

Running Prime 95 SFFT - 4 threads on 2 physical cores - on my 8700 pushes package power to 55W @ 4,3Ghz, and that's an AVX load which the Pentium doesn't even support.
 

birdie

Member
Jan 12, 2019
98
71
51
First IceLake based laptops are likely to be released this summer:

fugokrlwcsh21.png